Vector-Based and Submodel-Based Methods for Calculating Statistics of Outputs
This model provides examples of two different methods which can be used to generate statistics for time histories. These two methods are:
1) Vector-based: the time history or time series is converted to a vector and then the built-in vector functions are used to calculate the statistics.
- Pros: No iteration required for calculation and no submodel needed.
- Cons: Do not obtain a distribution for the time history values. As a result, need additional manipulation to determine percentile values which are not the median (i.e. the median is 50th percentile and can be determined fairly easily with this method, but the 72nd percentile would be relatively more difficult to determine with this method). Also, the time series is stored in a vector in addition to being stored in a time history output which requires more memory.
2) Submodel and Distribution Output: A static time, Monte Carlo submodel is created that uses the number of realizations equal to the number of time intervals in the regular interval time series. Each model realization is used to move forward through the time history by one time interval. The values for each realization are used in the submodel to create a distribution of values. The submodel output interface can then be set-up to output the distribution for time history values. The distribution provides the statistics of the time series.
- Pros: provides a distribution for the time history values which GoldSim calculates. Also, may have smaller memory requirements relative to the vector approach.
- Cons: Requires iteration through the realizations. The submodel set-up and input and output interface configuration may be conceptually more involved.